radv/gfx10: launch 2 compute waves per CU before going onto the next CU

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

diff --git a/src/amd/vulkan/radv_pipeline.c b/src/amd/vulkan/radv_pipeline.c
index efb94cdcd238f24d009322ffe31334ec441d8780..a58b0d6d00616b22f7b5f0af3f18b59f5c0c521f 100644 (file)
--- a/src/amd/vulkan/radv_pipeline.c
+++ b/src/amd/vulkan/radv_pipeline.c
@@ -4378,6 +4378,7 @@ radv_compute_generate_pm4(struct radv_pipeline *pipeline)
 {
        struct radv_shader_variant *compute_shader;
        struct radv_device *device = pipeline->device;
+       unsigned threads_per_threadgroup;
        unsigned threadgroups_per_cu = 1;
        unsigned waves_per_threadgroup;
        unsigned max_waves_per_sh = 0;
@@ -4402,10 +4403,14 @@ radv_compute_generate_pm4(struct radv_pipeline *pipeline)
                          S_00B860_WAVESIZE(pipeline->scratch_bytes_per_wave >> 10));
 
        /* Calculate best compute resource limits. */
-       waves_per_threadgroup =
-               DIV_ROUND_UP(compute_shader->info.cs.block_size[0] *
-                            compute_shader->info.cs.block_size[1] *
-                            compute_shader->info.cs.block_size[2], 64);
+       threads_per_threadgroup = compute_shader->info.cs.block_size[0] *
+                                 compute_shader->info.cs.block_size[1] *
+                                 compute_shader->info.cs.block_size[2];
+       waves_per_threadgroup = DIV_ROUND_UP(threads_per_threadgroup, 64);
+
+       if (device->physical_device->rad_info.chip_class >= GFX10 &&
+           waves_per_threadgroup == 1)
+               threadgroups_per_cu = 2;
 
        radeon_set_sh_reg(&pipeline->cs, R_00B854_COMPUTE_RESOURCE_LIMITS,
                          ac_get_compute_resource_limits(&device->physical_device->rad_info,